Your voice never leaves the machine
Speech recognition runs locally with whisper.cpp. No audio upload, no account, no telemetry, ever. The kind of privacy compliance teams actually require.
Quill turns your voice into text in any app, and your voice never leaves your PC. No account. No subscription. No telemetry. Pay once.
Free tier forever · Pro is a one-time purchase · works on an airplane
"um send the report to the team by friday new line thanks"
Types straight into the app you're already using
Why Quill
Speech recognition runs locally with whisper.cpp. No audio upload, no account, no telemetry, ever. The kind of privacy compliance teams actually require.
Buy it and keep it. No monthly rent on your own keyboard.
On a plane, in an air-gapped office, on hotel Wi-Fi, it just works.
One-click CUDA runtime for near-instant transcription on NVIDIA GPUs, with automatic CPU fallback.
Multilingual models with auto-detect, from tiny to large-v3-turbo.
Teach it your names, jargon, and product terms so it spells them right.
Push-to-talk while you hold the key, or tap to start and stop. Your call.
A floating widget with a live waveform and a flowing gradient ring, themeable to any accent, with normal, compact, and mic-only modes. It even floats above fullscreen games.
How it works
Anywhere, your email, an IDE, a chat box. The overlay appears without stealing focus.
Say "new line", "comma", "open paren" and they become real characters. Pause and it auto-stops.
Clean text is typed straight into the focused window, your clipboard stays untouched.
See it in action
Every transcript is saved locally. Press ≡ to search and revisit your history.
Make it yours
The difference
| Quill | Cloud dictation | |
|---|---|---|
| Audio stays on your device | ✓ Always | ✕ Streamed to servers |
| Works with no internet | ✓ | ✕ |
| Account required | No | Yes |
| Pricing | Pay once | Monthly forever |
| Telemetry by default | None | Usually |
| Usable in locked-down / compliance environments | ✓ | Rarely |
Built for the people who can't stream their voice to a third party, legal, medical, therapy, finance, and anyone who simply won't.
Pricing
The honest starting point.
Everything. Forever. No required internet connection to use.
For locked-down orgs.
A subscription costs more every year than Quill costs once.
FAQ
No. All speech recognition runs on your machine via whisper.cpp. After the one-time download of the app and a model, Quill needs no internet at all. There is no telemetry by default.
Any app with a text field, editors, browsers, chat apps, email, IDEs, even in-game chat. It types real keystrokes into whatever window has focus, so the clipboard stays untouched.
No. It runs on CPU out of the box. If you have an NVIDIA GPU, a one-click install enables CUDA for much faster transcription, with automatic CPU fallback.
Yes. Pro is a one-time purchase with a year of updates and a perpetual-use fallback license. No recurring charges to keep using what you bought.
Windows today. The architecture is local-first and Windows-native for the best latency and app compatibility.
Paste your license key into Settings → License. It's verified locally, no sign-in, no phone-home, consistent with the no-account promise.
Download Quill for Windows and dictate into any app, fully offline.
No account · No telemetry · Works offline